DESCRIPTION (Applicant's abstract): The purpose of the proposed study is to examine the interrelated influences of individual and contextual factors on adolescent alcohol, tobacco, and other drug use. We use Bronfenbrenner's ecology of human development theory as the overarching conceptual framework. Because of the long-standing centrality accorded to peers in adolescent substance use, we focus in particular on the peer network context and organize our investigation around the interrelationships of peer network factors with individual, family, and neighborhood characteristics. Our focus on peers is appropriate both because of their presumed importance in adolescent substance use and because of the critical need for advances in how peer influence is conceptualized and measured. We address this need through use of social network analysis methods, which provide an innovative but largely under-exploited means for advancing understanding of how peer context shapes adolescent drug use. The study's four specific aims are to: 1) use social network methods to identify and characterize the peer networks of adolescents and describe how the relationship between peer network characteristics and substance use changes over time; 2) determine whether peer network characteristics mediate the relationships between adolescent, family, and neighborhood characteristics and adolescent substance use trajectories; 3) determine whether adolescent, family, and neighborhood characteristics buffer or exacerbate the effects of peer network characteristics on adolescent substance use trajectories; and 4) determine whether the mediation and moderation models identified are specific or common to different substance use behaviors. We will address the study aims through a panel study of two cohorts of youth enrolled as they enter the 6th and 8th grades of the Winston-Salem/Forsyth County, North Carolina school system and followed over 3 years. Data will be collected from youth at school twice yearly. Data also will be collected yearly from parents by telephone to measure family and neighborhood characteristics. These data will be combined with data from archival sources regarding neighborhood characteristics. The study design will allow us to use a cohort sequential design in our analysis, thereby enabling us to model trajectories of substance use for youth from grade 6 through grade 10. Hierarchical Linear Modeling and Latent Curve Analysis will be used to test the study hypotheses.